Have I Been Trained is a new tool that lets artists know if certain copyrighted images have been used in the training of artificial intelligences.
This online tool, which relies on two large databases, was eagerly awaited at a time when complaints of copyright infringement are on the rise.
AI training, a problem for artists
Regularly, when large waves of data piracy take place, Internet users are redirected to the Have I Been Pwned tool to find out whether or not the confidentiality of their contact details has been compromised. Today, it is the artists who have to deal with this problem. With the explosion in popularity of image-creating AIs such as
Stable Diffusion or Midjourney, it becomes difficult to know with which images they were trained.
This is an additional difficulty for content creators, who have a hard time knowing whether or not their works have been used, and to what extent this concerns copyright. Indeed, current AIs do not create from scratch and are trained with gigantic databases available on the Web.
Artists can now turn to Have I Been Trained, which works on the same principle as Have I Been Pwned. The tool, already very popular, consults two large databases named LAION 5B and LAION 400M, which contain 5 billion and 400 million images. Artists can thus find out what is going on and possibly request the withdrawal of the works used.
Who was Have I Been Trained designed for?
First of all, it should be noted that the managers of Have I Been Trained work in close collaboration with LAION. These databases textually describe a huge volume of images, allowing AIs to make increasingly relevant associations. Concretely, the users of the tool can ask him to do a search by text or by image, after which they can observe the results and launch procedures if necessary.
” We are in partnership with LAION, who assembled these databases in order to ensure that future models [d’IA] are not trained with works that have been removed », Explains the platform.
Have I Been Trained, if it can be used by everyone, is above all reserved for artists, in response to several controversies and complaints filed. Getty Images and several artists have already turned against Stable Diffusion and Midjourney for illegally using a large number of copyrighted images. Of course, this won’t solve all the problems associated with using AIs, but Have I Been Trained seems able to save artists valuable time.
Sources: Have I Been Trained, ABC News
3